← All hypotheses

Dual-Vector vs Single-Vector Transduction

STOCHASTIC MODEL

STRC is too big for one AAV. The coding sequence is 5,325 base pairs. AAV packaging limit is 4,700. That is why Iranfar et al. (2026) use a dual-vector approach: split the gene into two halves, package in two separate AAV particles, inject both, hope they recombine inside the same cell.

Hope is the key word. For dual-vector to work, both viruses must enter the same cell. Then the two halves must find each other and recombine. Each step has its own probability. The losses multiply.

Our mini-STRC (residues 700-1775, 3,228 bp) fits in a single AAV. One particle. One transduction event. No recombination needed. This is not an incremental improvement. The math shows it is a different category.

The Math

Viral particle uptake follows a gamma-Poisson (negative binomial) distribution. Unlike simple Poisson, this accounts for the extreme cell-to-cell heterogeneity in viral delivery: cells near the injection site receive orders of magnitude more virus than distant cells. The dispersion parameter k captures this overdispersion.

For dual-vector: both AAV-A and AAV-B must enter the same cell. Each vector gets half the total dose. Heterogeneity helps dual-vector in mice (well-transduced cells get both), but helps less as overall MOI drops in larger human cochleas. Co-entry probability is computed analytically from the gamma-Poisson joint distribution.

Even after co-entry, the two halves must recombine intracellularly. We show results at R = 50% recombination efficiency (mid-range of published 30-100% for optimized trans-splicing designs). The gap widens at lower R.

Calibration Against Experimental Data

Two experimental data points from Omichi et al. (2020): single AAV2 achieved 83.9% OHC transduction, dual AAV2 achieved 65.6% functional co-transduction in mice at 3.75×1012 GC/mL. A simple Poisson model predicts only 35.9% dual (wrong). Fitting both points simultaneously yields a gamma-Poisson with λ = 8.1, k = 0.73, confirming extreme delivery heterogeneity (variance 11× the mean).

Omichi et al. (2020). Cochlear gene therapy with dual-AAV. PMC7270144

Human Prediction

Single-vector (mini-STRC)
Effective MOI
λ = 5.0 (gamma-distributed)
Transduction probability
77.8%
OHC with functional protein
9,336 of 12,000
Dual-vector (full STRC)
Effective MOI
λ/2 = 2.5 per vector
Transduction probability
27.3%
OHC with functional protein
3,278 of 12,000
Single-vector advantage
2.8×

At standard clinical titer (3.75×1012 GC/mL), single-vector mini-STRC delivers functional protein to 2.8 times more hair cells than dual-vector full STRC (at R = 50% recombination). At R = 30%, the gap widens to 4.7×. The advantage grows with scale: in the smaller mouse cochlea it is only 1.3×, but human cochlear geometry amplifies it.

Titer Dependence

The gap between single and dual widens at clinically realistic titers. Only at extreme titers (>1013 GC/mL, difficult to manufacture) does dual approach single efficiency.

Viral titerSingle-vectorDual-vectorGap
10¹⁰ 1.3% 0% +1.3%
3×10¹⁰ 3.8% 0% +3.8%
10¹¹ 11.5% 0.4% +11.1%
3×10¹¹ 27.2% 2.5% +24.7%
10¹² 53.1% 11.1% +42%
3.75×10¹² ← clinical 77.8% 27.3% +50.5%
10¹³ 88.5% 37.3% +51.2%
3×10¹³ 94.7% 44% +50.7%
📄 Full model: gamma-Poisson statistics, calibrated from Omichi 2020 View on GitHub: dual_vs_single_vector.py
Model uses gamma-Poisson (negative binomial) distribution calibrated from both experimental data points (Omichi 2020: 83.9% single, 65.6% dual). This accounts for delivery heterogeneity that simple Poisson ignores. Human predictions scale mouse parameters for cochlear volume (191 vs 2.5 μL perilymph), injection volume (15 vs 1 μL), and OHC count (12,000 vs 2,500). The dispersion parameter k is assumed equal between species (similar spiral geometry). Dual-vector numbers shown at R = 50% recombination efficiency.
← All hypotheses